STaR
A method that enables language models to improve their reasoning through self-generated chain-of-thought training. The model uses the reasoning traces from correctly solved problems as supervision signals for iterative self-improvement. It is a representative approach applying imitation learning to LLM reasoning fine-tuning.
Timeline (last 90 days)
RISED's self-distillation mechanism follows the same line of thinking as iterative self-improvement methods such as STaR and ReST, with the core idea of guiding the model's own progress using its successful examples.
STaR (Self-Taught Reasoner) and various distillation schemes are representative examples of imitation learning methods.
All Facts (2)
RISED's self-distillation mechanism follows the same line of thinking as iterative self-improvement methods such as STaR and ReST, with the core idea of guiding the model's own progress using its successful examples.
50%UnverifiedSTaR (Self-Taught Reasoner) and various distillation schemes are representative examples of imitation learning methods.
50%